Background The reverse algorithm for syphilis diagnosis consists of a treponemal antibody screening immunoassay followed by confirmatory nontreponemal antibody testing. It is increasingly used in the United States despite studies suggesting limited cost-effectiveness in high-prevalence groups. Methods In this retrospective cross-sectional study, we included men who have sex with men tested with the reverse algorithm in an Alabama HIV clinic between March 2015 and February 2017. Trep-Sure enzyme immunoassay (EIA) was used for the initial screen, followed by reflex nontreponemal reactive rapid plasma reagin (RPR) testing of specimens with positive results. Sociodemographic and clinical data were extracted from the electronic medical record and stratified according to EIA screen positivity. Quantitative EIA antibody index values were collected to assess test performance at various thresholds. Results Among 1693 men tested for syphilis with the reverse algorithm in HIV clinic, 808 (48%) had a positive initial EIA screen. A majority (53%) of men with subsequent RPR testing had a nonreactive RPR (EIA+/RPR-), and 19% (19/98) of these EIA+/RPR- samples tested had a negative confirmatory Treponema pallidum particle agglutination testing result. Analysis of quantitative EIA index values using a receiver operating characteristics curve suggested that a threshold >8 (rather the current threshold of antibody index 1.2) improved the performance of the test. Conclusions Among men who have sex with men tested in HIV clinic, the syphilis reverse algorithm was inefficient because of high rates of prior syphilis and false-positive EIA screening. Frequent syphilis screening in high-prevalence populations is an important part of the US epidemic response, and the traditional algorithm is preferred.